Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!samsung!daemon From: Stephen Smoliar Newsgroups: comp.music Subject: RE: MUSIC AND AI Message-ID: <25951@samsung.samsung.com> Date: 26 Jun 91 00:40:07 GMT Sender: daemon@samsung.COM Lines: 81 In article <3191@lee.SEAS.UCLA.EDU> bsmith@turing.seas.ucla.edu (Brian Smith) writes: >In article <10936@idunno.Princeton.EDU> eliot@phoenix.Princeton.EDU (Eliot >Handelman) writes: >> >>It's much more interesting to focus on the listener, because apprehension >>is the simplest act of creativity. > >Listening is pretty fascinating stuff, but I think that composition is a much >more interesting problem since one can't compose without listening. Where have you been for the last forty years, Brian? Following the Second World War, there was no end of experimentation in music; and much of which emerged had nothing to do with listening (at least at the time of composition). The early days of computer-synthesized sounds provide a good case in point. Composers often labored long and hard to debug the theoretical formulation of what they wanted yet rarely had much intuition regarding what the tape would sound like when all the processing was done. You may wish to take the ethical position that one OUGHT NOT to compose without listening, but do not expect all practicing composers to accept that position. > To some >extent, everyone is a composer (i.e. humming arbitrary tunes, melodic contours >in speech, etc.), so it does seem to be a worthwhile area of study. I think you are homing in on an important point here, Brian; and I would like to try to push it a bit further. What you are really talking about here is BEHAVIOR, and one of my favorite hobby-horses is that there is more to behavior than can be captured in logical calculi or neural nets. The trouble is that we to not do a terribly good job when it comes to DESCRIBING such behavior. The sorts of protocol analyses which were performed by Newell and Simon were little more than self-fulfilling prophecies--descriptions based on a foundation of symbol manipulation which they assumed HAD to be there. Music, on the other hand, does not lend itself to such symbol-based descriptions because, as Ed Hall has been suggesting on comp.music, the actual PRACTICE of MAKING MUSIC has precious little to do with the symbols of music notation. Getting a MACHINE to "make music" (i.e. to exhibit such behavior) may thus be viewed as a major challenge to artificial intelligence, because it is an aspect of behavior which has been ignored (and certainly not accounted for) by most of the progress in AI to date. > Once the >knwoledge gained through listening is captured, how do we use it in the >performance domain? > This is another example of how thinking about music should force us to expand the current horizons of artificial intelligence. Much of the artificial intelligence community seems inclined to live in a world in which "learning" is a matter of adding declarative sentences to some kind of "knowledge base." However, the above sentence captures an element of learning which is much truer to behavior as we know it: How one behaves "in the performance domain" is a reflection of what has happened during past listening experiences. This is not a new idea Brian. Quite some time ago, Minsky wrote a wonderful essay on the role of a musical composition AS TEACHER. Unfortunately, we have made precious little progress in implementing any of these ideas. Nevertheless, those ideas still deserve more attention. After reading Volume 47 of ARTIFICIAL INTELLIGENCE, we should be at least SKEPTICAL about what logical calculi can ultimately offer us. At the very least, we should be encouraged to work on problems which logic does not "fit" as comfortably as it does is, for example, choosing the parameters for the design of an elevator system. What IS "knowledge gained through listening?" We really do not have the slightest idea? We do not yet even have a handle on how we know that what we are hearing NOW is the same tune we heard five minutes ago! So far I have only been able to pursue such questions as peripheral activities, but my current hunch is that SERIOUS attention to these matters may ultimately lead to a significant shift in our current paradigms for artificial intelligence. =============================================================================== Stephen W. Smoliar Institute of Systems Science National University of Singapore Heng Mui Keng Terrace, Kent Ridge SINGAPORE 0511 BITNET: ISSSSM@NUSVM "He was of Lord Essex's opinion, 'rather to go an hundred miles to speak with one wise man, than five miles to see a fair town.'"--Boswell on Johnson